Copyright © 2001, GaAs MANTECH, Inc.
InGaP Makes HBT Reliability a Non-Issue
Aditya Gupta, Alex Young and Burhan Bayraktaroglu
ANADIGICS, Inc, 35 Technology Drive, Warren, NJ 07059
Tel: (908)668-5000
agupta@anadigics.com
ABSTRACT
Among the many benefits InGaP brought to
the GaAs-based heterojunction bipolar transistor
(HBT) over the last few years, nothing is more
profound than the drastic improvement it
provided in the long-term reliability. Today, it is
relatively common to obtain activation energies
higher than 1.5eV with InGaP HBTs operating
under bias and temperature conditions suitable
for modern telecommunication applications. This
significant improvement in reliability coupled
with excellent fabrication control afforded by
selective etching has made InGaP HBT the prime
choice for large volume fabrication of cellular
phone power amplifiers. All new HBT fabrication
facilities around the world today either have or
plan to have 6-inch InGaP HBT wafers as their
only choice for power amplifiers.
INTRODUCTION
There are a number of fundamental differences in
the construction of heterojuctions made from InGaP
compared to older approaches using AlGaAs. First,
InGaP emitter layers do not have the type of defects
associated with Al-containing layers such as AlGaAs.
Second, the energy band gap alignment in
InGaP/GaAs heterointerface favors the use of abrupt
junctions compared to graded junctions commonly
employed with AlGaAs. Low interface defects
combined with better confinement of holes at the
interface, are some of the reasons for achieving
consistent predicted lifetimes in excess of 109 hours
with InGaP HBTs. Previous investigations have
consistently shown that InGaP emitter Heterojunction
Bipolar Transistors (HBTs) have reliability superior
to AlGaAs emitter HBTs [1-6].
Using MOCVD growth techniques, the excellent
reliability advantages of InGaP HBTs have been
extended to 6-inch wafers. It has been shown recently
that the mean-time-to-failure (MTTF) of the devices
in 6-inch manufacturing fabrication process is in
excess of 2x109 hours with an activation energy of EA
= 1.8 ± 0.2 eV. A lifetime of 3250 hours has been
measured at a junction temperature, Tj, of 275°C and
current density, J C= 25 kA/cm2 [7].
As GaAs HBT technology matured over the last
decade, a significant and steady progress has been
made in reliability improvements. Most of these
improvements can be attributed to superior material
quality available nowadays with MOCVD. However,
device fabrication is also a major factor in achieving
a reliable final device. As the technology matured,
the assessment of reliability was often confused
between the factors introduced by details of the
fabrication method and the materials growth. This is
particularly the case with AlGaAs HBTs, whose
emitter-base heterojunction must be graded for
proper device operation. The preparation of this
junction is an art that is practiced in a unique way by
each material supplier. To make the matters worse,
each device manufacturer practices the fabrication of
AlGaAs HBTs with significant variations. In sum, the
constituents of reliability limiting factors are not
clearly assessed. This situation is much less
ambiguous with InGaP HBTs. Because of minimal
conduction band offset between InGaP emitter and
GaAs base regions, an abrupt junction is used. Such a
junction is better able to confine holes at the
heterojunction and therefore offer reduced gain
sensitivity to temperature. Further, an abrupt junction
when coupled with excellent etch selectivity,
provides the manufacturability control needed for
high volume applications. It is this repeatability and
reproducibility of manufacturing process that allows
us to investigate the effects of epitaxial material on
the reliability of the HBT devices, independent of the
fabrication process. The material contributions are
significant as epitaxial material suppliers continue to
develop their 150mm growth capabilities. We will
examine the impact of epitaxial material on the
reliability of InGaP HBTs fabricated in a high
volume 150mm wafer facility. We will discuss our
results obtained with traditional and other rapid
assessment techniques.
DEVICE FABRICTION
The devices studied in this work consist of
MOCVD grown InGaP HBTs supplied by
commercial epitaxial material vendors on 150mm
GaAs substrates. The device structure includes a
500Å InGaP emitter and a 1000Å thick base layer,
carbon doped at NA = 4 x 1019 cm-3. Reliability test
devices, shown in Figure 1, were fabricated with a
four finger 3x20µm2 emitter area HBT using the
ANADIGICS, Inc. production fabrication process.
The reliability test structure includes a 5Ω ballast
resistor on each emitter finger. The process also
includes a 1.0 µm ledge structure made from the
depleted InGaP layer. Wafers are thinned to 100 µm
and via contacts are used to ground the emitter.
Devices are then assembled into packages using a
AuSn solder die attach process to reduce the thermal
resistance.
methods. In the more conventional temperature
accelerated tests, devices were stressed at 25kA/cm2
current density and Vce=3.3V at three elevated
temperatures (275°C, 300°C, and 325°C). DC current
gain was continuously monitored in-situ and the
failure criteria for this test was defined as a 20%
degradation in gain. The data was captured using an
automated measurement system that logs the
collector current once per hour for each device. The
current acceleration technique relies on selecting a
suitable high current density and junction
temperature to rapidly assess the propensity of the
device to degrade. This test only yields a pass-fail
type assessment and no clear activation energy is
determined.
0.10
C
ol
le
ct
or
C
ur
re
nt
(A
0.08
Ib = 500 µA
Ib = 400 µA
0.06
I = 300 µA
0.04
b
I = 200 µA
b
0.02
I = 100 µA
b
0
0
1
2
3
4
5
Vce (V)
Figure 2: Common emitter I-V characteristics of
an InGaP emitter HBT.
Figure 1: Reliability test device. The DC
characteristics of this device is shown in Figure 2.
Typical DC current gain of this device is 100 with
small signal microwave performance of ft = 28 GHz
and fmax = 40 GHz.
RELIABILITY TEST RESULTS
Device reliability assessment was conducted using
both temperature and the current acceleration
A typical plot of the normalized DC Beta over the
lifetest at Tj = 300°C is shown in Fig. 3. Similar to
previous reports, we observe two degradation
mechanisms in this material [2]. The first mechanism
results in a small gradual decrease in the collector
current that is not catastrophic. This accounts for
approximately a 10% reduction in the DC current
gain. Device failure occurs when the current drops
rapidly as a result of a second degradation
mechanism. A failed device exhibits an increased
base current. Figure 4 shows the Gummel
measurement taken both before the device was
stressed and after a 1058 hour stress at Tj = 300°C.
Both measurements were made at an ambient room
temperature of TA=25°C. As shown in this figure, the
degradation in current gain is due to increased base
current similar to previous observations [2]. The
magnitude of change in base current, however, is
strongly dependent on Vbe value. At Vbe less than
1.2V, the current gain rapidly approaches unity.
1.2
Using the time at which 50% of the devices
reached failure, t50, the failure time at each
temperature is plotted in Fig. 6. A straight line fit to
these data yields a MTTF = 3 x 109 hours at Tj = 125
°C with an activation energy of EA = 1.8 eV for our
production process. These values of MTTF and EA
are consistent with other reported results of InGaP
HBTs. This fact demonstrates that the InGaP emitter
HBT is a robust device on 150mm substrates as well
as other substrates.
T = 300 C
j
Normalized DC Beta
1.0
0.8
0.6
0.4
0.2
0
1
10
100
1000
10 4
Time (hr)
MTTF = 3 x 109 hr
@ T = 125 C
109
Fig. 3: Normalized DC Beta in-situ measurements during
300 ºC lifetest.
j
Current (A)
10 -1
MTTF (hr)
7
Before Stress
After Stress
105
103
IC
10 -3
10
E = 1.8 eV
A
10
-5
1
10
1.25
IB
1.5
1.75
2
2.25
2.5
1000/Tj (1/K)
10 -7
10
-9
0.6
0.8
1
1.2
1.4
1.6
Fig. 6: Median-time-to-failure (MTTF) as a function of
inverse junction temperature, Tj , at JC= 25 kA/cm2 with
VCE = 3.3 V.
Vbe (V)
Fig. 4: Gummel plot measured before (solid lines) and after
(dashed lines) 1058 hour stress at Tj = 300 °C.
1.0
Tj = 325 C
0.8
109
0.6
275 C
0.4
0.2
0
10
Vendor A
300 C
MTTF (hr)
% Cumulative failures (x100)
The cumulative failures are plotted over the
lifetest for each temperature. At Tj = 325°C, there are
some infant failures as demonstrated in Fig. 5.
The dependence on the material quality has been
demonstrated in our evaluation of InGaP emitter
HBT epitaxial material from three different suppliers.
Each vendor has grown InGaP HBT structures with
identical characteristics of Beta, Rb, BVceo and Vbe.
Given the mature process technology and the stability
of our production fabrication process, we believe the
process does not contribute to any differences in the
reliability results from different vendors.
Vendor B
107
Vendor C
105
103
100
1000
10 4
Time (hr)
Fig. 5: Cumulative failures plotted for Tj = 275, 300, and
325 °C.
101
1.25
1.5
1.75
2
2.25
2.5
1000/Tj (1/K)
Fig. 7: Comparison of MTTF for three epi material
vendors.
The results of life tests (see Figure 7) on each set
of material indicates that the activation energy varies
from 1.5 eV to greater than 2.0 eV for the different
vendors. The MTTF at Tj = 125 °C is >10 8 hours
even at the lowest activation energy
150kA, 1.5V Stress
1.E+00
1.E-01
1.E-02
CURRENT ACCELERATED TESTS
150kA, 1.5V Stress
1.E-01
1.E-02
Current
1.E-03
1.E-04
20hr
3hr
1.E-05
1.E-07
1.E-04
1.E-05
1.E-06
1.E-07
20hr
3hr
1hr
5min
1.E-08
1.E-09
0.75
1
1.25
1.5
Vbe (V)
Figure 9: High current stress has negligible base current
increase on reliable devices.
the resulting statistics are used to assess the overall
reliability of the wafer. MTTF and activation energy
are not determined by this screening method.
SUMMARY
We have demonstrated the reliability of our
production InGaP emitter HBT fabrication process
using 150 mm GaAs substrates. The MTTF was
determined to be 3 x 109 hours at Tj = 125 °C with an
activation energy of EA = 1.8 eV at a collector current
density of J C= 25 kA/cm2 with VCE = 3.3 V. We have
also demonstrated the influence of epitaxial material
growth on the reliability of 150 mm InGaP emitter
HBTs. We have obtained lifetimes that range from 1
to 3 x 109 hours at Tj = 125 °C with activation
energies that range from 1.5 to greater than 2.0 eV.
ACKNOWLEDGEMENTS
The authors would like to thank the ANADIGICS,
Inc. wafer fabrication personnel that do a great job
processing wafers.
1.E+00
1.E-06
Current
1.E-03
The assessment of HBT reliability by temperature
accelerated life tests is a well known and proven
technique. A major drawback of this method,
however, is the significant effort required in
packaging the devices and the long time taken to
complete the tests and obtain the desired data. At
ANADIGICS, HBT reliability is also assessed by a
technique that does not require any packaging and
yields results in a few hours. In this method, typically
performed at the wafer level, the reliability test
device (Figure-1) is stressed for a brief period (e.g.
one hour) at extremely high current densities
(150kA/cm2 in our test) and the Gummel plot is
monitored before and after the test. The HBT is
stressed both due to the high current density as
described in [8] and due to the high Tj as a result of
self heating. In our tests, Tj is estimate to be 300ºC
under VCE=1.5V and JC=150kA/cm2. Our experience
shows that HBTs with a propensity to degrade will do
during this “hammer” test and that 1 hour is adequate
to show this tendency (Figure-8). Degraded HBTs
show increased base current and base ideality factor
as in conventional life tests. Thus, we believe , the
same failure mechanism is activated by this method.
Reliable devices (Figure-9) will not show any
significant degradation in β after this stress. This test
is performed on a sample of HBTs on the wafer and
1hr
5min
1.E-08
1.E-09
0.75
1
1.25
1.5
Vbe (V)
Figure 8: Current accelerated degradation showing increase
in base current similar to temperature accelerated
degradation.
REFERENCES
[1] Low, T.S., et al., 1998 GaAs IC Symposium Technical
Digest, pp. 153-6.
[2] Yeats, B., et al., 2000 GaAs MANTECH Digest, pp. 131-5.
[3] Ueda, O., Microelectronics Reliability, 1999, vol. 39, pp.
1839-55.
[4] Pan, N., et al., IEEE Electron Device Letters, April 1998, vol.
19, no.4, pp 115-7.
[5] Oka, T., et al., 2000 GaAs MANTECH Digest, pp. 137-40.
[6] Alderstein, M., et al., IEEE Trans. Electron Devices, Feb.
2000,vol. 47, no. 2, pp. 434-9.
[7] Cheskis, D., et al, 2000 GaAs Reliability Workshop, Seattle,
WA
[8] Sugahara, H., et al, GaAs ’93 Technical Digest, pp. 115-18.